Forecasting and future work - Severe Convective Storms and Tornadoes: Observations and Dynamics - page 421

sum of d of both forecast and unforecast events), the critical success ratio (CSI),
which is a measure of the relative number of forecast events to the sum of all the
forecast, unforecast, and non-events, and the average lead time, which is the
duration of time between when the forecast is issued and when the forecast event
begins, if it does.
One may also compute the average value of the differences between forecast
variables (precipitation amount, largest hail size, strongest wind speed). These
quantities, however, may be misleading indicators of forecast skill. Suppose that a
forecast is perfect, save for an error in location. In other words, the afore-
mentioned empirical measures of forecast error would be poor, but if the forecast
were shifted in space by a small distance, they could be significantly reduced. Then
it is the phase of the forecast that is in error. It is somewhat of an art to devise
indicators of forecast accuracy.
Ed Lorenz suggested in the 1960s that errors in numerical forecasts should
grow more rapidly as the initial estimate of atmospheric variables is improved and
smaller and smaller scales are resolved. The predictability of convection may not
be much greater than the time it takes air parcels to enter a storm and exit it,
which is the advective time scale. If so, then there could be a theoretical limit that
is relatively short. Errors on very small scales are amplified by nonlinear terms
and show up on resolved larger scales. On the other hand, some convective
systems like supercells persist for time periods longer than the advective time
scale, which could result in extended predictability. This extended predictability
could be caused by a reduced effect of nonlinear advection terms. It was noted
earlier, for example, how some nonlinear advection terms are small when there is
low helicity.
7.2 FORECASTING AND CLIMATE CHANGE
In the last decade, when temperatures averaged globally have been near record
highs (at least with respect to the last century or so), and in the aftermath of a
spring with record tornado activity in the U. S. (2011), the topic of how climate
change will affect the climatology of severe convective storms has become of
greater interest to the public, in part as a result of enhanced coverage by the
media. In particular, global climate models have been run which have included the
physical process of enhanced radiative forcing due to increased greenhouse gases.
Two of the ingredients associated with supercells, CAPE and vertical shear, can be
measured from the very long-range forecasts from these models to determine
whether these ingredients will be made more or less conducive to supercell forma-
tion. It is recognized that these climate models do not explicitly represent
convection and that many processes are parameterized using methods that have
varying degrees of accuracy, some not very accurate.
It is thought that low-level moisture will increase as low-level temperature
increases, which, if the temperature aloft does not warm appreciably, will result in
greater CAPE. However, it is also thought that the overall pole-to-equator tem-